Fault Location Identification in Smart Distribution Networks With Distributed Generation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Fault Location Identification in Smart Distribution

Networks with Distributed Generation


Jose Cordova M. Omar Faruque
Department of Electrical & Computer Engineering Department of Electrical & Computer Engineering
Center for Advanced Power Systems (CAPS) Center for Advanced Power Systems (CAPS)
Florida State University (FSU), Tallahassee, FL, USA. Florida State University (FSU), Tallahassee, FL, USA.
jdc13b@my.fsu.edu faruque@caps.fsu.edu

Abstract—In this paper, based on the measurements available


from the smart grid devices such as Advanced Metering smart equipment, use of these techniques for exact location
Infrastructure (AMI), Reclosers, Distributed Generators (DG) identification became close to reality.
and other IEDs, a fault location identification method is
proposed that can accurately identify the fault location. The II. STATE-OF-THE-ART FAULT LOCATING TECHNIQUES
algorithm is suitable for distribution networks with DG and A. Current Fault Locating Methods
smart measurement infrastructure that can transmit event-
driven data such as pre and post-fault voltages or currents of Significant research [3-5] has been done to identify the
scarce number of meters. A MATLAB-based state estimation fault location in transmission systems, where the system is
(SE) technique is applied to identify the fault location, and the balanced and employs SCADA data to monitor it. Fault
open source tool OpenDSS is used to perform the offline locating in the distribution network is more difficult due to its
validation by creating faults in simulation. The IEEE 37 node unbalanced nature and the lack of sufficient measurements.
test feeder was modified to add DGs at different locations and Smart grid implementation is going to change this in the near
was used for the validation of the proposed algorithm. Multiple future. Fault locating methods that rely on the characteristic
faults (both symmetrical and unsymmetrical) were created for
response of the electrical network, such as the Fault Current
the exhaustive validation of the technique. In more than 90% of
the cases the location was identified in the first guess and for the Based method, which relies on the recording data of previous
remaining 10% it took second guess and the faulted node was fault events (fault current and location) so that they can be
indeed near to the first guess. compared with the response of the system when an actual
fault occurs [3]. These methods have low cost but suffers
Index Terms--Distributed power generation, Fault currents, from inaccuracy [4].
Fault location, Power distribution faults, Smart grids, State Another technique [5], called the Impedance Method,
Estimation relies on the measurements at the substation of the
distribution network. Generally, utilities have accurate
I. INTRODUCTION metering devices at different locations for power flow
Distribution lines are constantly subject to several fault information. The meters measure voltage and current at the
conditions caused by factors such as adverse weather, car feeder when a fault occurs, thus a “fault impedance” value
accidents, contact with animals, and equipment malfunction can be calculated during a fault. This impedance value is
[1]. Fault currents can damage the distribution infrastructure proportional to the distance where the fault is located, giving
and may create energy interruption that could be momentary some idea of where the fault could be. However, when a
or sustained depending on the magnitude of the fault current distribution network has many laterals and is populated with
and nature of the fault. In order to minimize outage time and DG sources, the accuracy of this method is questionable.
provide high quality service, it is crucial for utilities to locate Other techniques such as Traveling Waves method [5] rely
the fault position as quickly as possible. In general, crew on powerful meter devices. The wave characteristic of
members look manually for blown fuses and circuit breakers alternate current (AC) allows this method to determine a fault
in order to identify the fault location [2]. The process often location when comparing the difference between normal
takes several hours. phase and the faulted phase of the traveling wave when a
The need for faster and more accurate fault locating fault event takes place. These methods rely on time
techniques has led to the development of some automated synchronization between meters generally performed by
fault location methodologies [2] that have not been widely Global Positioning Satellite (GPS) systems. These techniques
used due in part to lack of sufficient monitoring points until can be very accurate after some filtering of the signal since
the recent increase in smart meter penetration. With the they can be too noisy. The disadvantage of implementing the
deployment of
Traveling Wave method is the large investment in a metering here will be tested for a three-phase, unbalanced and
infrastructure across the distribution network. multilaterals distribution network. Also, it will consider
Another method [5], known as Signal Injection technique, multiple sources of DG with their current contribution during
is based on injecting a particular frequency signal through fault. In order to reduce the cost of data collection
buses-to-ground circuit expecting it to return through earth infrastructure, the potential of collecting data from the
along the fault line by ground point. The fault line is located already installed smart meters and other intelligent electronic
by identifying the section where the signal flows through. devices are discussed.
Although the metering structure is not as powerful as the III. FAULT EVENTS ASSESSMENT WITH SMART GRID
GPS-based method used by the Traveling Wave method, the TECHNOLOGY
Signal Injection technique still requires a massive amount of
signal detectors such as mobile detectors, instrument One fundamental problem for the assessment of fault
transformers Hall Sensor technology, remote detectors, etc. events in distribution networks is the lack of data given the
[6]. This will reflect in a large investment by the utilities. scarce metering points in the system. However, the smart grid
It can be inferred that most of the fault locating methods technologies are spreading and could overcome this problem
used in the transmission side cannot be implemented in the and would be useful for fault analysis. Modern equipment
distribution system given the unbalanced topology of such as Reclosers, AMI, IED, transfer switches and other
distribution networks and the large cost of the installation of devices could provide necessary measurements for a fault
complex metering devices for the application of the methods location analysis. Some of the more advanced Recloser
mentioned above. controllers [13] include a communication application for
Fault locating methods for distribution networks that remote control and monitoring, fault indication, power quality
assume a radial structure of main feeders without considering analyzing and automation in medium-voltage secondary
multilateral topology are presented in [7-9]. Javadian et al. distribution systems. These devices can store data for a
[10] presents an algorithm that separates the distribution sequence of events that can be accessed remotely. Similarly,
network into radial sections having their own measurements. data from Advanced Metering Infrastructure (AMI), a group
This method needs a significant change in the topology of the of technologies that provides connections between meters at
system by adding meters at precise locations. Bretas and the consumers’ end and the utilities company can be used.
Salim [11] presented a method using the positive-sequence Besides the energy consumption of the customer, smart
apparent impedance. This assumption oversimplifies the meters can notify the utilities when an outage occurs (by
problem since the method does not take into account the using event-driven data capture) and can record power quality
inherent distribution-side unbalances and mutual phase events [14]. One commercial example is the GE family of
coupling. Reference [12] shows a method that compares kV2c meters that have programmable sag and swell logging
fault-on current with the pre-fault current obtained in an capabilities that can be triggered in case of fault conditions
iterative method to determine the fault location. Although the [15].
method was tested for a three phase system, it did not A fault location technique based on state estimation is
consider multilateral multisource topology. presented in the following sections. The state estimation needs
fault-on currents of the sources, and fault-on voltages profile
of some buses that can be gathered from a metering
B. Fault Locating Techniques in Networks with DG. infrastructure and delivered to an operation center of the
One of the major challenges for fault locating algorithms utilities in order to determine fault locations. Fig. 1 presents
to be effective is the inclusion of DG in the distribution how the AMI and other intelligent electronic devices (IED)
system. The most commonly present DGs are the can be connected to a server that can gather information for a
Photovoltaic system (PV) and the Wind Turbines (WT). Centralized Fault Management System. The data obtained by
Distributed generators are generally located close to the the meters can be processed in order to have a real-time
particular load that is intended to be served. During the fault, monitoring of fault location identification. In this paper, the
distributed generators become another source of contribution objective is to present a reliable fault locating method that can
use the field measurements of a distribution network and
to fault currents. A distribution network with PVs and WTs
locate the fault in seconds on the screen of the operator.
behaves differently than one with no DG presence during a
fault event. Fault current also depends on the location of the
DGs, their nominal voltage value and the specific control
strategy of the DG present. Fault location identification for
networks with DGs becomes more complex and therefore it is
a field of high interest.

From the discussion above, it can be seen that a method is


needed that can be applied to an unbalanced and multi-source
system. Moreover, the controlled limiting current scheme is Figure 1. Smart Grid Communication StructureState Estimation (SE) in
an important part of DGs behavior during a fault and it will Distribution Networks
be taken into account in this paper. The method presented
A. State Estimation (SE) General Equation Based on these assumptions, pseudo-currents can be added
SE is a powerful tool whenever measurement points are for equation (2) as follows:
scarce in a system. This mathematical method is based on v1abc (F)
finding an accurate regression of every possible state in the
system (i.e. voltages and currents across a network). This p
vabc (F)
______ v1abc (F)
technique is widely used in transmission networks and now its grid
use is increasing in the distribution side. The SE principle is Iabc (F) Φ
________
= viabc (F) . (3)
simple and it is calculated by the following equation: 0
z̃ =Ax (1) vnabc (F)
i
Iabc (F)
The variable is the column vector of m number of
measurements. Matrix A is a characteristic matrix of the 0
i
system formed by a permutation matrix Φ, and the Ybus matrix. Where Iabc (F) is the returning fault current at node i and
1
The variable is the column vector known as state variables Iabc (F) is the substation current seen by the feeder. Since all
which are the unknown values in the network. Solving for load currents are neglected during fault conditions we have:
would give a profile of the estimated state variables of the i grid
Iabc F =-Iabc (F) (4)
measured points and the non-measured points in the network.
The expanded form of (1) is given in (2). Variables on the left- Equation (4) is valid for SE when the distribution system
hand side are the voltage and current measurements during a has only one source (the grid). However, with the presence of
fault condition. In order to be able to find a solution for the DG the situation changes and SE algorithm needs to take care
state variables, more than half of the nodes in the system have of fault currents injected by the DGs.
to be measured.
C. Fault Current Contribution of Distributed Generators in
v1abc (F) Distribution Network.
p
vabc (F) v1abc (F)
Distributed Generation is spreading, and hence adding
______ more and more generation sources to the grid. In order to
1 Φ
________
Iabc (F) = viabc (F) . (2) develop a fault locating algorithm for this changing
i
circumstances in a distribution network, all the fault current
Iabc (F) vnabc (F) contributions of distributed generators must be taken into
n account. Neglecting load currents, the fault current at node i
Iabc (F)
becomes:
It can be seen that the missing states for which there is no fault,i
measurement available, must be estimated from the rest of the Iabc F =- (5)
information of the system available through measurements. Where IGrid F , Isolar F , Iwind F and Iothers (F) are the
This is called State Estimation and it is a process of estimating contributed fault currents by the main substation, solar, wind,
unknown states from measured quantities. A good estimate and other types of DG, respectively. It is expected that the
will leave out small errors, and compensate missing data by fault current data for all the DGs are available from the
giving the “best estimate” of the state of the entire system with measurements of the sites.
the measurements available.
D. State Estimation in Distribution Network with the
B. State Estimation in Distribution Networks with no DG presence of PV systems only.
When using SE, it is a common practice to make analytical Equation (3) shown in the previous section can be
assumptions in order to “fix” measurements to a certain value. modified to include all the pseudo-currents and current
These theoretical values are called pseudo-measurements and measurements from all the sources. Assuming a system has
if valid assumptions are made, they make the SE more connection to a main substation and a number of photovoltaic
accurate. In distribution networks, the common current (PV) systems as sources we have:
measurement available is at the main feeder circuit breaker at
the substation. Therefore, to obtain the rest of the current v1abc (F)
measurements, the following assumptions are made to perform p
vabc (F)
SE: ______
CB,1
Iabc (F)
- The load currents are negligible in comparison to the fault
v1abc (F)
current. 0
Φ
________
- For ground faults, fault currents are considered to be PV,1 = viabc (F) . (6)
Iabc (F)
equal to the source current, expecting that it will return vnabc (F)
through ground. PV,k
Iabc (F)
- In the line-to-line fault scenario, the measurement at the fault,i
-Iabc (F)
feeder substation provides the value of the returning
current through the lines. 0
,
Where is the current contribution of the PVs the fault lasts more than a specified time, a trip could be
present, and (5) becomes issued to disconnect the PV unless other protection devices
fault,i PV,1 PV,2 PV,k trip first. During current limiting mode, the inverter outer
Iabc F = F + Iabc F + Iabc F + … + Iabc (F) (7) power control loop breaks and it effectively turns into a
Equation (7) defines the fault current Iabc F at node i as
fault,i current source. The fault response of an inverter is dominated
the sum of all the currents of the PV systems connected to the by its method of current limiting. This should be reflected in
network plus the contribution from the grid. the development of a fault model for the inverter [17].
Therefore, the PV systems were simulated as a current source
IV. FAULT LOCATING ALGORITHM IN DISTRIBUTION in OpenDSS. The OpenDSS model behaves like a current
NETWORKS BASED ON STATE ESTIMATION source with a simple interface inverter. When a Vmin value is
Let us assume the total number of sources in the system is reached at the grid side of the PV, such as when a fault
s. The source bus number position is defined in a vector p as: occurs, the PV system injects constant current. In this study,
the current injected during a fault is considered as 1.25 per
p = 1, 2, … , ,…, (8) unit of rated current.
Following the notation in (8), the current contributions
during a fault from every source become Iabc F , Iabc F …
Iabc F .

An iterative program in MATLAB was developed for


locating the fault. The current measurement at node i becomes
if a source is present. At the k-th iteration, a
hypothetical faulted node introduces the value of ,
which is calculated with either (4) or (5) depending if there are
one or more sources. All other node currents are neglected.
Iabc F if i = pj
Iiabc (F)= fault
Iabc F if i = k (9)
0 otherwise
After each iteration, the regression of the state estimation
is compared to see if the hypothetical fault node k gives an
accurate state of the system. The estimated voltage profile can
be compared to the measurements to see if the predicted states
are approximate with the following error calculation:
k= z̃ [k]-Ax[k] (10)
Figure 2. IEEE 37 node test feeder with 5 PV plants added
Where · is the Euclidean norm. Vector z̃ [k] and x are the
measurement and estimated values vectors for all the nodes in TABLE I. PV SPECIFICATIONS
the system, respectively when the returning fault current is set
at the k-th hypothetical node in the k-th iteration. Therefore, AC Power Power
Phases
AC Voltage Vmin Controller
(kW) Factor (kV) (p.u.) Strategy
the fault is located by the iteration that suits the best to the Constant kW,
fault current hypothetical returning node. The node yielding 350 1 3 0.48 0.8
kVAR
the smallest error is considered as the faulted node.
TABLE II. SPECIFICATIONS OF THE TRANSFORMER FOR PV
V. TESTING & RESULTS CONNECTIONS
For validating the algorithm, the IEEE 37 test feeder kVA kV-high kV-low % reactance
has been used. The test feeder does not have any DGs on it. 350 4.8, Delta 0.48, Delta 3.0
For our case studies, we have added five PV plants at buses
711, 731, 729, 718 and 720 through transformers. The faults
were simulated using OpenDSS and pre-fault and post-fault B. Measurement Locations
data were used in the state estimation algorithm implemented Pegoraro et al. [18] presents an algorithm for metering
in MATLAB. location optimization that can be used with the state
estimation method presented on this paper. For this case study,
A. IEEE 37 Nodes Test Feeder with Distributed Generation
current measurements are taken from the main feeder at the
Different types of faults were simulated in OpenDSS with substation and at the locations of PV. Voltage measurements
the IEEE 37 nodes test feeder [16]. Five PV generators were are taken from the end of the branches resembling a customer
placed as shown in Fig. 2. Identical PV systems and metering point. A total number of 24 meters are used on the
transformers were used and their specifications are shown in IEEE 37 nodes test feeder and their functions are shown in
Table I and II, respectively. table III.
Generally, during a fault in a PV system, the control
TABLE III. NUMBER OF METERS IN THE TEST FEEDER
function limits the fault current to a threshold of 1.2-1.4 p.u. If
Element
Number of Voltage Current For the grounded transformer connection, Table VI
Meters Measurement Measurement presents detailed data for the two nodes that were not
Main Substation 1 Yes Yes identified in the first guess. For each of these cases, the
PV 5 Yes Yes relative error between the actual faulted node and the best
guess is very close, meaning that the state estimator was very
Bus 18 Yes No close to identifying the faulted node correctly. Table VII
C. Fault simulations and Results shows the identification error for grounded and ungrounded
connections, that is, the difference between the SE of the
The current and voltage profiles resembling field correct faulted and the first guess output of the algorithm.
measurements during fault conditions were collected using These errors could be due to the unknown load currents. The
OpenDSS and then fed to the state estimator. The SE analyzes state estimator runs with the assumption that the returning
the fault-on measurements and determines an estimate of the current is equal to the current contribution of all sources
entire system when placing the fault at a hypothetical faulted present in the system which neglects load currents.
node. Accuracy of the estimation is measured by its error and
it determines the faulted node. Overall, the algorithm shows a 90% of accuracy in finding
the right faulted node for a two-line-to-ground fault scenario.
Three-phase Balanced Fault
A three-phase balanced bolted fault was set in thirty four TABLE IV. TWO-LINE-TO-GROUND FAULT LOCATING ALGORITHM
OUTPUT
nodes of the system, one at a time. Fault location algorithm
presented in section V was used to determine the location of First
Second
the fault event. Connection Guess Failed Total
Guess
(Correct)
Thirty four nodes, one node at a time, were set to a 3-phase Delta-Wye 31 2 1 34
balanced fault. After feeding the fault-on voltages and currents
Delta-Delta 29 3 2 34
the measurements to the state estimator, fault locations were
guessed accurately for each fault-on condition. It has been
found that the state estimator accurately locates all thirty four TABLE V. DELTA-DELTA CONFIGURATION FAULT LOCATING SECOND
GUESSINGS
fault locations as a first guess which translates to 100%
accuracy in identifying fault locations performance. 1st Guess 2nd Guess
Faulted Identified Node Identified Node Identification
Further tests were made by changing the IEEE 37 nodes Node SE SE Error (%)
Node Node
test feeder main substation transformer configuration. The test Error Error
feeder as given in [16] is totally ungrounded at the 4.8 kV-side 729 733 12.2567 729 12.2787 0.18
hence giving the zero sequence current no returning path. New
738 737 9.8942 738 9.8962 0.02
configuration was changed from a Delta-Delta connection as
to a Delta-Y with the neutral solidly grounded. This change 732 775 13.5940 732 13.6232 0.21
was made in order to determine the algorithm accuracy in both
grounded and ungrounded systems. Since the faults were set TABLE VI. DELTA-WYE CONFIGURATION FAULT LOCATING SECOND
on the 4.8 kV side of the system, changing the 230 kV GUESSINGS
transformer connections did not give any effect in algorithm 1st Guess Identified 2nd Guess
accuracy. All thirty four nodes subject to a fault were correctly Faulted Node Identified Node Identification
guessed. The technique presented a 100% success rate for Node
Node SE Error Node SE Error
Error (%)
three-phase balanced fault.
729 733 12.2567 729 12.2787 0.18
Two-phase-to-ground Fault
738 737 9.8942 738 9.89624 0.02
A two-phase-to-ground fault is applied to the same 34
nodes as in the previous case. The setup of the system was the TABLE VII. FAILED NODES ERROR FOR TWO-LINE-TO-GROUND FAULT
same as in the previous case except for the fault type. Table IV
presents the output of the algorithm containing the number of 1st Guess
Correct Identification
Connection Identified
cases where the fault was identified in the first guess, the Faulted
Node
Error (%)
number of nodes located in the second guess and the nodes
Delta-Wye 718 733 0.889391
that failed to be identified. It also shows the output for
grounded and ungrounded transformer configuration. Twenty 718 775 0.243743
nine faulted nodes were guessed correctly for the ungrounded Delta-Delta
741 775 0.517613
connection and thirty for the grounded configuration.
Table V shows the nodes that were located as a second
best guess with their respective state estimation error in the Line-to-Line Fault
ungrounded transformer connection. It can be seen that the Table VIII presents the output of the algorithm for a line-
algorithm failed to guess them correctly in the first guess by a to-line fault applied to the IEEE 37 nodes test feeder.
very small error (less than 0.2%). Grounded and ungrounded transformers connections presented
similar results not guessing the correct fault location of two
nodes by a very small percentage error as presented in table faults and transformer connections to demonstrate that the
IX and table X. The algorithm missed by a small percentage. technique can be implemented in the distribution network with
Again, it can be seen that the algorithm guessed the right fault different configurations. Although in real-life it is expected
location for a 90% of the nodes that were set to a line-to-line that fault current contribution of DG will be available from the
fault. field measurement provided by the smart grid, in this case an
off-line simulation program (OpenDSS) is used to collect
measurement data. The important aspect of this technique is
TABLE VIII. LINE-TO-LINE FAULT LOCATING ALGORITHM OUTPUT FOR
that it can identify the fault location with more than 90%
DELTA-DELTA CONFIGURATION accuracy in the first guess even with the presence of DGs in
the network. For very small number of cases, the fault location
First Guess was found in the second best guess with a small error
Connection Second Guess Failed Total
(Correct)
difference which can be attributed to the approximations and
Delta-Wye 30 2 2 34
assumptions that were made. The SE error in these cases was
Delta-Delta 29 2 3 34 very small indicating that just for a fractional value the guess
was incorrect. By guessing the node closest to the fault, the
TABLE IX. DELTA-DELTA CONFIGURATION FAULT LOCATING SECOND algorithm can be used together with technique in [18] to
GUESSINGS determine the location of the fault in a line section.
1st Guess Identified 2nd Guess REFERENCES
Faulted Node Identified Node Identification
Node SE Error (%) [1] T. Baldwin, D. Kelle, J. Cordova, N. Beneby, “Fault locating in
Node SE Error Node Distribution Networks with the Aid of the Advanced
Error
Measurement Infrastructure,” Power Systems Conference at
729 733 12.2567 729 12.278 0.18
Clemson University, 2014.
738 737 9.89424 738 9.8962 0.02 [2] “IEEE Guide for Determining Fault Location on AC
Transmission and Distribution Lines”IEEE Standard C37.114-
2014.
TABLE X. DELTA-WYE CONFIGURATION FAULT LOCATING SECOND [3] G. C. Lampley, “Fault analysis on electrical distribution system,”
GUESSINGS in IEEE Rural Electric Power Conference, 2008.
1st Guess 2nd Guess [4] “Fault detection and location on electrical distribution,” in Rural
Faulted Identified Node Identified Node Identification Electric Power Conference. IEEE 2002.
Node Error (%) [5] (5)Y. Li, Q. Du, X. Qi, Q. Pang, G. Zhu, “A review of Single-
SE
Node Node SE Error Phase-to-Ground Fault Location Methods in Distribution
Error
Networks,” 4th International Conference on Electric Utility
729 733 12.256 729 12.2787 0.17 Deregulation and Restructuring and Power Technologies, 2011.
[6] G. Buigues, V. Valverde, I. Zamora, J. Mazon, and E. Torres,
738 737 9.894 738 9.8962 0.02
“Signal Injection techniques for fault location in distribution
networks.” International Conference on Renewable Energies and
Power Quality, 2012.
Single-line-to-ground Fault [7] Y. Lu, L. Hua, J.Wu, G.Wu, and G. Xu, “A study on effect of
dispersed generator capacity on power system protection,” IEEE
Table XI presents the results for single-line-to-ground fault Power Eng. Soc. Gen. Meeting, Tampa, FL, Jun. 2007.
tests for the test feeder. From a total of thirty four nodes that [8] Z. Xiangjun, K. Li, W. Chan, and S. Sheng, “Multi-agents based
were subject to a fault, twenty five resulted as the correct protection for distributed generation systems,” in , Proc. IEEE Int.
guess in the first outcome of the algorithm. Seven were Conf. Electric Utility Deregulation, Restructuring and Power
selected as the second best guess, and only two failed to be Technologies, Apr. 2004, vol. 1, pp. 393–397.
[9] S. Javadian, A. Nasrabadi, M.-R. Haghifam, and J. Rezvantalab,
identified. It was observed that even though the algorithm “Determining fault’s type and accurate location in distribution
guessed the correct faulted node as a second guess, the first systems with DG using MLP neural networks,” in Proc. Int. Conj.
guess were adjacent to the correct ones. This implies that the Clean Electrical Power, Jun. 2009, pp. 284-289.
state estimation can still give a good starting point for the [10] S. Javadian, M. Haghifam, and N. Rezaei, “A fault location and
utilities technical crews to look for the fault location. protection scheme for distribution systems in presence of dg using
mlp neural networks,” in Proc. IEEE Power Energy Soc. Gen.
Meeting, Jul. 2009, pp. 1–8.
TABLE XI. SINGLE-LINE-TO-GROUND FAULT LOCATING RESULTS [11] A. Bretas and R. Salim, “Fault location in unbalanced DG
First Guess systems using the positive sequence apparent impedance,” in
Identification Second Guess Failed Total Proc. IEEE/Power Eng. Soc. Transmission Distribution Conf.
(Correct)
Expo.: Latin America, Aug. 2006, pp. 1–6.
Quantity 25 7 2 34 [12] D. Johnsonbaugh and A. Girgis, “Fault location for distribution
systems with distributed generation using a modified three phase
method,” presented at the Power Systems Conf. Distributed
VI. CONCLUSIONS Generation, Advanced Metering and Communication, Clemson,
SC, Mar. 2004.
In this paper, an algorithm is proposed that can be used to [13] ABB, “Grid Automation, Recloser Protection and Control
identify the fault location based on limited measurement data RER615, Product Guide,” p. 13, Sep 2013.
available from the smart equipment that is already becoming [14] S. Depuru, L.Wang, and V. Devabhaktuni, “Smart meters for
ubiquitous in a distribution network. To demonstrate the power grid: Challenges, issues, advantages and status,”
Renewable and Sustainable Energy Reviews, vol. 15, no. 6, pp.
effectiveness of the proposed algorithm, the IEEE 37 node test 2736–2742, 2011.
feeder was used. The method was tested for different types of
[15] GE Energy, “Commercial & Industrial Electricity Metering,” pp.
1–8, May 2012.
[16] IEEE PES Distribution Systems Analysis Subcommittee Radial
Test Feeder, http://ewh.ieee.org/soc/pes/dsacom/testfeeders/
[17] C. Plet, T. Green, “Fault Response of Inverter Interfaced
Distributed Generators in Grid-Connected Applications”, Electric
Power Systems Research 106 (2014) 21– 28.
[18] S. M. Brahma, “Fault Location in Power Distribution System with
Penetration of Distributed Generation,” IEEE Transactions on
Power Delivery, Vol. 26, No. 3, July 2011, 1545-1553.

You might also like